Epidemiology/Health Services/Psychosocial Research

O R I G I N A L A R T I C L E

Weight Change and Duration of

Overweight and Obesity in the Incidence
of Type 2 Diabetes
S. GOYA WANNAMETHEE, PHD besity is well recognized as an impor-
A. GERALD SHAPER, FRCP
O tant risk factor for type 2 diabetes
(1–3), and weight control has been
proposed as a means of preventing type 2
diabetes (4). Obesity enhances insulin resis-
tance (5), a condition characterized by
OBJECTIVE — To examine the relationship between weight change and duration of over- increased insulin production and impaired
weight and obesity and the incidence of type 2 diabetes in a cohort of middle-aged British men.
glucose intolerance (6), both of which are
RESEARCH DESIGN AND METHODS — We carried out a prospective study of car- reversible with weight loss (7,8). It may
diovascular disease in men aged 40–59 years at screening (1978–1980), drawn from one gen- therefore be expected that weight loss and
eral practice in 24 British towns, who completed a postal questionnaire 5 years later (Q5) and prevention of weight gain are beneficial for
for whom data on BMI at year 1 (Q1) and Q5 were available (n = 7,100). Men with diabetes at the primary prevention of type 2 diabetes.
Q1 or Q5 and men with hyperglycemia at Q1 were excluded from the study (n = 184). The Weight gain has been associated with an
main outcome measure was type 2 diabetes (physician-diagnosed) during a mean follow-up increased risk of type 2 diabetes in both
period of 12 years starting at Q5 (1983–1985). men and women in several U.S. popula-
tions (9–13). Although intervention studies
RESULTS — In the 6,916 men with no history or evidence of diabetes, there were 237 inci- in very high-risk subjects (e.g., severely
dent cases of type 2 diabetes during the mean follow-up period of 12 years, a rate of 3.2/1,000
person-years. Substantial weight gain (.10%) was associated with a significant increase in risk
obese subjects, obese subjects with a family
of type 2 diabetes compared with that in men with stable weight (relative risk [RR] 1.61 [95% history of type 2 diabetes, or subjects with
CI 1.01–2.56]) after adjustment for age, initial BMI, and other risk factors. Excluding men who impaired glucose intolerance) have sug-
developed diabetes within 4 years after the period of weight change increased the risk further gested that weight reduction significantly
(1.81 [1.09–3.00]). After adjustment and exclusion of men who developed diabetes early in the reduces the risk of type 2 diabetes (14–16),
follow-up, weight loss ($4%) was associated with a reduction in the risk of type 2 diabetes, the benefits of weight loss have been less
compared with that in the stable group, that reached marginal significance (0.65 [0.42–1.03], consistent in population-based settings.
P = 0.07). A test for trend that fitted weight change as a continuous covariate showed the risk Although weight loss has been shown to be
of diabetes to increase significantly from maximum weight loss to maximum weight gain associated with a reduction in the incidence
(P = 0.0009). The lower risk associated with weight loss was seen in obese ($28 kg/m2) and of diabetes in women (12), other studies
nonobese subjects and in men with normal (,6.1 mmol/l) and high ($6.1 mmol/l) nonfasting
blood glucose levels. Although not statistically significant, this is consistent with a benefit from
have found little or no benefit (14). Indeed,
weight loss. Risk of type 2 diabetes increased progressively and significantly with increasing lev- some have observed an increased risk of
els of initial BMI and also with the duration of overweight and obesity (P , 0.0001). diabetes (17,18), and other studies have
suggested that weight fluctuation (weight
CONCLUSIONS — This study confirms the critical importance of overweight and obesity, loss followed by weight gain) may be dia-
particularly of long duration, in the development of type 2 diabetes. The data support current betogenic (19). This article examines the
public health recommendations to reduce the risk of type 2 diabetes by preventing weight gain relationship between weight change during
in middle-aged men who are not overweight and by encouraging weight loss in overweight and a 5-year period and the risk of type 2 dia-
obese men. betes during the subsequent 12 years in a
large prospective study of .7,000 men.
Diabetes Care 22:1266–1272, 1999

RESEARCH DESIGN AND

METHODS — The British Regional
Heart Study is a large prospective study of
cardiovascular disease consisting of 7,735
men aged 40–59 years selected from the
From the Department of Primary Care and Population Sciences, The Royal Free and University College Med- age-sex registers of one group general prac-
ical School, London, U.K.
Address correspondence and reprint requests to Dr. S. Goya Wannamethee, Department of Primary Care tice from 24 towns in England, Wales, and
and Population Sciences, Royal Free and University College Medical School, Rowland Hill Street, London Scotland. The criteria for selecting the town,
NW3 2PF, U.K. the general practice, the subjects, and the
Received for publication 3 February 1999 and accepted in revised form 14 April 1999. methods of data collection have been
Abbreviations: CHD, coronary heart disease; NHANES, National Health and Nutrition Examination Sur-
vey; Q1, year 1; Q5, year 5; RR, relative risk. reported (20). Research nurses adminis-
A table elsewhere in this issue shows conventional and Système International (SI) units and conversion tered a standard questionnaire to each man
factors for many substances. that included questions on smoking habits,

1266 DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999


alcohol intake, physical activity, and med-
ical history. Several physical measurements
were made, and nonfasting blood samples
were taken to measure biochemical and
hemotological variables including blood glu-
cose. Glucose concentration was measured
in serum with an automated analyzer
(Technicon SMA 12/60; Technicon, Tarry-
town, NY). Diurnal variation in glucose
concentration was modest, with a peak
trough difference of 0.4 mmol/l (21).
Details of classification methods for smok-
ing status, physical activity, and BMI have
been reported (20,22). Five years after the
initial examination (Q5) (1983–1985), a
postal questionnaire similar to the one
administered at screening was sent to all
surviving men to obtain information on
medical history and changes in smoking,
drinking behavior, and weight. The ques-
tionnaire was completed by 98% of avail-
able survivors (n = 7,262).
BMI
At initial screening, weight and height were
measured, and BMI was calculated with
weight/height2 used as an index of relative
weight. Five years later (Q5), the men were
asked to state their weight in pounds or
kilograms, and BMI was calculated for each
man based on their reported weight and on
measured height at the initial screening.
“Lean BMI” was defined as ,25 kg/m2,
“overweight” was defined as a BMI of
25.0–27.9 kg/m2, and “obese” was defined
as a BMI $28 kg/m2, which represents the
upper quintile of the BMI distribution in all
men at Q1. These cutoffs broadly corre-
spond to the definition of overweight and
obesity in the U.K., where $25 kg/m2 is
considered to be overweight, 28–30 kg/m2
is considered to be bordering on obesity,
and $30 kg/m2 is considered to be obese
(23). BMI data were available at both Q1
and Q5 in 7,100 men.
Smoking
From the combined information at screen-
ing and Q5, the men were classified as
having never smoked cigarettes, long-term
former smokers (former cigarette smokers
at both Q1 and Q5), recent former smok-
ers (former cigarette smokers at Q5 only),
and current cigarette smokers at Q5 who
were categorized into three groups (1–19,
20, and $21 cigarettes/day).
Physical activity
At the initial screening, the men were asked
to indicate their usual pattern of physical
DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999
activity, which included regular walking or
cycling, recreational activity, and sporting
activity. A physical activity score was derived
for each man based on frequency and type
of activity, and the men were grouped into
six broad categories based on their total
score: none, occasional, light, moderate,
moderately vigorous, and vigorous (22).
Recall of physician diagnosis
At screening and at Q5, the men were
asked to recall whether they had ever been
told by a physician that they had any of the
12 major specified conditions listed on the
questionnaire: ischemic heart disease
(angina, heart attack, coronary thrombosis,
or myocardial infarction), “other heart trou-
ble,” high blood pressure, stroke, gout, dia-
betes, gall bladder disease, thyroid disease,
arthritis, bronchitis, asthma, or peptic
ulcer. Subject recall of coronary heart dis-
ease (CHD), stroke, and angina has been
validated with general practitioners’ record
reviews, and these have shown to have
good agreement (24,25).
Measures of weight change
An index for weight change was deter-
mined for each man by calculating the per-
centage change in body weight since the
initial screening (26). For a man of average
weight (76 kg), a gain of 3.0 kg (4%) con-
stituted a change in weight. Subjects with
weight loss were defined as men who had
lost at least 4% of body weight; weight
gain was defined as those who had gained
.4% of body weight. Those who had
gained or lost ,4% of body weight were
classified as stable. The men were grouped
into four weight-change categories: weight
loss, stable, gain of 4–10% of body weight,
and gain of .10% in body weight.
Follow-up
From initial screening, all men have been fol-
lowed up for all cause mortality, cardiovas-
cular morbidity, and the development of type
2 diabetes up to December 1995, a mean
period of 16.8 years (range 15.5–18.0 years)
(27). Information on death was collected
through the established “tagging” procedures
provided by the National Health Service reg-
isters in Southport (for England and Wales)
and Edinburgh (for Scotland). New cases of
type 2 diabetes were ascertained via a postal
questionnaire sent to the men at Q5 for
each individual, systematic reviews of pri-
mary care records in 1990 and 1992, a fur-
ther questionnaire to 6,483 surviving
members of the cohort residing in Britain in
Wannamethee and Shaper
1992, and a review of all death certificates
for any mention of diabetes. The question-
naire at Q5 had a response rate of 98%, and
the 1992 questionnaire had a response rate
of 91%. A diagnosis of diabetes was not
accepted on the basis of questionnaire data
unless it was confirmed in the primary care
records. All men with a history of diabetes
at screening or at the Q5 questionnaires and
men with nonfasting blood glucose levels
$11.1 µmol/l at screening were excluded
from the analyses (n = 184). This report
studied only the 7,100 men who completed
the Q5 questionnaire and specifically the
6,916 men who did not have diabetes at
screening or on the Q5 questionnaire or
hyperglycemia at screening. Data on fol-
low-up since the Q5 questionnaire are pre-
sented for an average follow-up of 11.8
years (range 10.5–13.0 years), and follow-
up has been achieved for 99% of the cohort.
Statistical methods
The Cox proportional hazards model (28)
was used to obtain the relative risks (RRs) for
the weight-change groups and BMI groups
adjusted for age, smoking, BMI, physical
activity, and recall of CHD and high blood
pressure. To obtain greater statistical power
in assessing the relationship between weight
change and risk of type 2 diabetes, weight
change was also fitted as a continuous
covariate in its original form. In other words,
the percentage of weight change in either
direction was used, with loss of weight hav-
ing a negative value and gain in weight hav-
ing a positive value. The purpose was to
determine the significance of the change in
risk over the continuum from maximum
weight loss to maximum weight gain.
RESULTS — In the 6,916 men with full
information on weight change and with no
history or evidence of type 2 diabetes, there
were 237 incident cases of diabetes during
the mean follow-up period of 12 years, a
rate of 3.2/1,000 person-years.
Most men (56%) were stable in weight,
31% gained weight, and 13% lost weight
(Table 1). One-third of the men who lost
weight were obese at Q1, but 5 years later,
only 15% remained obese. Of the men who
gained a substantial amount of weight
(.10%), the proportion who were obese
increased threefold during the 5 years.
Weight change and risk of type 2
diabetes
Table 2 shows the age-adjusted rate per
1,000 person-years and adjusted RR by the
1267
Weight change, duration of overweight, and diabetes

Table 1—Weight-change categories with mean weight change and percentage obese at Q1
and Q5
Weight change n (%)
Loss ($4%) 937 (13)
Stable 3,844 (56)
Gain
4–10% 1,673 (24)
.10% 462 (7)
weight-change categories. Risk was in-
creased only in the men who had gained a
substantial amount of weight (.10%),
although the difference from the stable
group was not significant (Table 2, column
A). However, initial weight (BMI) was high-
est in those who had lost weight and low-
est in those who had gained .10% weight
(26.88 vs. 23.91 kg/m2). Adjustment for
initial body weight reduced the risk in the
weight-loss group and increased the risk in
the weight-gain groups so that the
increased risk in the substantial weight-
gain group was now significant (Table 2,
column B). A test for trend over the con-
tinuum of weight change from maximum
weight loss to maximum weight gain was
significant (P = 0.0005).
Weight loss has shown to be associated
with smoking status, diagnosis of CHD,
high blood pressure, and physical activity
(26,29,30), all of which are potential risk
factors for type 2 diabetes (1,2,31). Further
adjustment for these factors reduced the
risk in the weight-loss group, although the
difference from the stable group was still not
statistically significant, possibly because of
small numbers (Table 2, column C). How-
ever, a test for trend on the continuum of
Obese ($28 kg/m2) (%)
Mean weight
change (kg) Q1 Q5
26.3 33 15
10.2 18 20
14.7 13 26
114.4 12 38
weight change from maximum weight loss
to maximum weight gain was significant (P
= 0.002). Because weight loss is often asso-
ciated with ill health and may precede the
diagnosis of diabetes, we excluded men
who died or developed diabetes within 4
years of follow-up from the Q5 question-
naire (Table 2, column D). Exclusion
reduced the risk in the weight-loss group to
marginal significance (P = 0.07) and
increased the risk in the weight-gain groups.
Weight gain of .10% remained associated
with a significant increase in the risk of dia-
betes. The trend from maximum weight
loss to maximum weight gain became more
significant (P = 0.0009).
Effect of initial BMI
In all men, the incidence of type 2 diabetes
increased progressively with increasing BMI
(P , 0.0001), and the risk was significantly
increased even at BMI levels of 25.0–27.9
kg/m2 (Table 3). The age-adjusted RRs and
95% CIs for the three groups (,25,
25.0–27.9, and $28 kg/m2) were 1.00, 2.24
(1.54–3.23), and 5.11 (3.60–7.28). We
examined the effects of weight loss by initial
BMI levels adjusted for age, smoking, phys-
ical activity, and recall of high blood pressure
and CHD. Within each weight-change cate-
gory, there was an increasing incidence rate
of diabetes with increasing BMI at screening.
Weight loss was associated with a modest
reduction in RR (27%) compared with the
stable group, although this difference was
not statistically significant, possibly because
of the small numbers involved. Substantial
weight gain (.10%) was associated with an
increase in risk of diabetes in men who were
lean or overweight at baseline. In the latter
group, the increase in risk was not signifi-
cant, presumably because of the small num-
bers involved. In subjects who were obese
($28 kg/m2) and already at high risk for
developing diabetes, further weight gain
made little difference in their RR of type 2
diabetes. Exclusion of men who developed
diabetes in the early follow-up made minor
differences in the levels of risk. A test for
trend on the continuum of weight change
from maximum weight loss to maximum
weight gain after these exclusions was
significant in the lowest BMI category (P =
0.004) and was of marginal significance in
the heavier BMI groups (P = 0.06 and P =
0.07 respectively).
Blood glucose, weight loss, and type 2
diabetes
Blood glucose concentration is a major risk
factor for type 2 diabetes (1,2,31). We exam-
ined the relationship between weight change
and risk of diabetes by separating men with
screening (nonfasting) glucose levels ,6.10
mmol/l (n = 5,531, 145 cases) and those
with levels $6.10 mmol/l, the top fifth of the
glucose distribution (n = 1,344, 92 cases).
Data on blood glucose were not available in
41 men, and these men have been excluded.
Men in the top fifth of the glucose distribu-
tion showed significantly higher rates than

Table 2—Age-adjusted rates per 1,000 person-years and adjusted RR for diabetes according to weight-change category

Age-adjusted rate
per 1,000 RR (95% CI)
Weight change person-years A B C D
Loss (.4%) 3.2 (31/937) 1.07 (0.72–1.58) 0.77 (0.52–1.15) 0.74 (0.49–1.10) 0.66 (0.41–1.04)*
Stable 3.0 (125/3,844) 1.00 1.00 1.00 1.00
Gain
4–10% 3.3 (60/1,673) 1.11 (0.81–1.52) 1.26 (0.93–1.72) 1.17 (0.85–1.60) 1.21 (0.86–1.70)
.10% 4.4 (21/462) 1.47 (0.93–2.33) 1.89 (1.19–3.01) 1.61 (1.01–2.56) 1.81 (1.09–3.00)
Test for trend-fitting — 0.24 0.0005 0.002 0.0009
weight change in its
original continuous form (P)
A, age-adjusted; B, age and BMI at screening; C, age, BMI at screening, smoking, physical activity, and recall of CHD and high blood pressure; D, as in C, exclud-
ing men who died or developed diabetes (n = 342) within 4 years of follow-up from Q5. Analysis was based on 6,574 men and 196 cases of diabetes. *P = 0.07.

1268 DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999

 Wannamethee and Shaper

Table 3—Age-adjusted diabetes rate per 1,000 person-years and adjusted RR of diabetes by initial BMI and weight-change categories

Weight change
Gain
Initial BMI (kg/m2) All men Loss ($4%) Stable 4–10% .10%
,25 (51/3,187)
Rate per 1,000 person-years 1.4 1.1 (3/262) 1.4 (26/680) 1.2 (12/930) 3.0 (10/315)
Adjusted RR 0.73 (0.22–2.42) 1.00 0.90 (0.45–1.81) 2.49 (1.15–5.41)
Adjusted RR* 0.29 (0.04–2.14) 1.00 0.90 (0.42–1.90) 3.02 (1.36–6.73)
Test for trend* P = 0.004
25–27.9 (84/2,444)
Rate per 1,000 person-years 3.1 2.7 (10/363) 2.8 (45/1,467) 4.1 (23/522) 6.4 (6/92)
Adjusted RR 0.91 (0.46–1.82) 1.00 1.32 (0.79–2.22) 1.96 (0.81–4.74)
Adjusted RR* 0.79 (0.45–1.78) 1.00 1.33 (0.74–2.36) 2.02 (0.76–5.32)
Test for trend* P = 0.06
$28 (102/1,285)
Rate per 1,000 person-years 7.0 5.6 (18/312) 7.4 (54/697) 11.0 (25/221) 9.4 (5/55)
Adjusted RR 0.67 (0.39–1.16) 1.00 1.29 (0.80–2.09) 0.95 (0.37–2.42)
Adjusted RR* 0.66 (0.36–1.19) 1.00 1.37 (0.82–2.30) 0.91 (0.32–2.59)
Test for trend* P = 0.07
Data are cases/n of men or RRs (95% CI) and are adjusted for age, smoking status at Q5, physical activity, and recall of CHD and hypertension. *Excluding men
who died or developed diabetes within 4 years of follow-up from Q5.

those with levels ,6.1 mmol/l (6.6 vs. baseline (i.e., for $5 years) and those who greater risk of diabetes than men who had
2.4/1,000 person-years, P , 0.0001). had become overweight, obese, or mark- been in the BMI category for ,5 years.
Weight loss was associated with a reduced edly obese during the last 5 years. Because Subjects who had been obese (28–29.9
risk in both blood glucose groups after of the focus on duration of obesity, all men kg/m2) for ,5 years showed a near three-
adjustment for age, initial BMI, physical who lost weight ($4%) were excluded fold increase in RR, and this increased to
activity, and recall of hypertension and CHD from this analysis. Table 4 shows the rela- more than fourfold in subjects who had
(RR 0.73 [95% CI 0.42–1.27] and 0.72 tionship between attained BMI at Q5 and been obese for $5 years or those who were
[0.40–1.31]), although the difference was risk of diabetes in men who had not lost markedly obese for ,5 years. Men with
not statistically significant, possibly because weight. The risk of diabetes increased with marked obesity ($30 kg/m2) for $5 years
of the small numbers involved. Modest and increasing BMI level and also with the had eight times the risk of developing dia-
substantial weight gains in subjects with duration of overweight or obesity, even betes compared with men who were not
glucose levels ,6.1 mmol/l were both asso- after full adjustment (test for trend across overweight (,25 kg/m2).
ciated with an increased risk of type 2 the seven BMI groups, P , 0.0001). At
diabetes (1.46 [0.99–2.23] and 2.39 each level of BMI $25.0 kg/m2, men who CONCLUSIONS — Obesity is an
[1.38–4.15], respectively). In subjects with had been overweight, obese, or markedly important and well-established risk factor
high glucose levels, neither moderate nor obese for $5 years consistently had a for diabetes. We have previously reported
substantial weight gain indicated an
increased risk (0.85 and 0.71, respectively).
A test for trend from maximum weight loss Table 4—Risk of diabetes by BMI at Q5 and duration of overweight and obesity
to maximum weight gain was significant in
those with glucose levels ,6.1 mmol/l (P ,
Cases Rate per 1,000 Age-adusted Adjusted
0.0001) but was not significant in subjects BMI at Q5 n of diabetes person-years RR RR (95% CI)
with higher glucose levels.
,25 2,386 34 1.3 1.00 1.00
Duration of obesity and risk of 25–27.9
type 2 diabetes ,5 years 623 17 2.5 1.80 1.74 (0.96–3.15)
In addition to level of obesity, the duration $5 years 1,541 50 3.0 2.17 2.25 (1.45–3.47)
of obesity has also been considered to be an 28–29.9
important risk factor for type 2 diabetes ,5 years 408 18 4.0 2.91 2.68 (1.50–4.81)
(32). The men were grouped according to $5 years 553 39 6.7 4.87 4.74 (2.99–7.51)
their attained BMI at Q5. Those who were $30
overweight (25.0–27.9 kg/m2), obese ,5 years 48 3 6.5 4.92 4.36 (1.33–14.28)
(28–29.9kg/m2), or markedly obese ($30 $5 years 355 42 11.8 8.66 8.04 (5.06–12.74)
kg/m2) were further divided into those who *Adjusted for age, smoking status at Q5, physical activity, and recall of CHD and hypertension. Men who
had already been in these BMI categories at have lost weight ($4%) have been excluded (n = 937).

DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999 1269

Weight change, duration of overweight, and diabetes
on the positive relationship between BMI at
baseline and the incidence of type 2 dia-
betes during 12 years of follow-up (31).
The present study, based on 12 years of fol-
low-up starting from Q5, confirms that the
risk of diabetes is significantly increased
even at levels of BMI (25.0–27.9 kg/m2)
that include the average weight of middle-
aged British men. The risk of diabetes
increases progressively with increasing
BMI. The increase in risk seen at BMI lev-
els of $25 kg/m2 has also been reported in
women (33).
Weight gain and type 2 diabetes
Substantial weight gain (.10%) was signifi-
cantly associated with an increased risk of
diabetes. The finding of increased risk of
diabetes with weight gain is consistent with
several other adult population studies in
both men and women (9–13). In the pres-
ent study, men with below-average BMI
who gained a substantial amount of weight
showed a significant threefold increase in
risk compared with subjects whose weight
remained stable, and the risk was twofold
(albeit not significant) in subjects who were
already overweight (BMI 25.0–27.9 kg/m2).
This is consistent with the findings that
weight gain is associated with an increase in
insulin resistance (34) and deterioration in
glucose tolerance (35), factors that are
strongly associated with the development of
diabetes. In subjects who were already
obese or who had high serum glucose lev-
els and thus already at high absolute risk of
developing diabetes, weight gain did not
appear to have any further adverse effect.
This suggests that the effects of weight gain
are mediated through an increase in glucose
levels or insulin resistance.
Duration of overweight or obesity was
also found to be a risk factor for diabetes as
observed in other studies (32). Compared
with men with BMI levels ,25.0 kg/m2,
the risk of diabetes increased nearly three-
fold in men who were moderately obese
(28.0–29.9 kg/m2) for ,5 years and nearly
fivefold in men who were moderately obese
for .5 years. In men with marked obesity
($30 kg/m2), the RR of diabetes was nearly
twice as great in men who had been
markedly obese for $5 years compared
with men whose marked obesity was of
shorter duration.
Weight loss and risk of diabetes
The hazards of obesity are well known and
include hypertension, dyslipidemia, glu-
cose intolerance, and insulin resistance
1270
(36,37). Weight reduction has been shown
to be associated with an improvement in
blood pressure, blood lipid profile, and
glucose intolerance (8,17). With weight
loss, insulin sensitivity increases, and even
modest amounts of weight loss can pro-
duce a significant improvement in glycosy-
lated hemoglobin levels (38,39). It might
therefore be expected that weight reduction
should diminish the risk of developing dia-
betes. Evidence for the benefits of weight
reduction comes from intervention studies
in subjects at very high risk for diabetes.
Weight loss in severely obese subjects
appears to deter the progression from
impaired glucose tolerance to diabetes (15).
Weight reduction has been shown to
significantly reduce the risk of diabetes in
obese subjects with a family history of dia-
betes (16). In the Malmo study in Sweden,
weight loss in persons with impaired glu-
cose tolerance was associated with a reduc-
tion in risk of developing diabetes (14). In
the few population studies that have exam-
ined weight change and the risk of dia-
betes, the beneficial effects of weight loss on
the risk of diabetes have been less consis-
tent. In some population studies, weight
loss appeared to have no benefit (13) or
was associated with increased risk (17,18),
and some studies have even suggested that
weight fluctuation (weight gain followed by
weight loss) may be diabetogenic (19).
In the present study, weight loss was
associated with a reduction in subsequent
risk of diabetes during a 12-year period.
The apparent benefit of weight loss was
seen in both obese and nonobese subjects
and in men with low and high glucose lev-
els. Although the reduction in risk in the
weight-loss group compared with the sta-
ble group was not statistically significant
when using grouped data, possibly due to
small numbers and lack of statistical power,
the trend for increasing risk on the contin-
uum of weight change from maximum
weight loss to maximum weight gain was
highly significant. This suggests that the
more weight one loses, the greater the
reduction in risk, and the more weight one
gains, the higher the risk. The reduction in
risk associated with weight loss is consis-
tent with that reported in the Nurses’
Health Study, a large prospective study of
.100,000 female nurses in which a signifi-
cant reduction in risk of diabetes with
weight loss was observed (12). Weight loss
in women with BMI .27 kg/m2 was asso-
ciated with a 30% reduction in the risk
of developing diabetes compared with
DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999
women whose weight did not change. The
magnitude of reduction was very similar to
our own findings in men with BMI $28
kg/m2. By contrast, the U.S. National
Health and Nutrition Examination Survey
(NHANES) showed little benefit from
weight loss except in the 25–29 kg/m2 BMI
group. Weight loss often precedes the onset
of diabetes, and diabetic subjects who were
diagnosed early in the study period had not
been excluded (13). In the present study,
exclusion of men who died or developed
diabetes within 4 years of weight loss
reduced the risk in the weight-loss group
further, and the trend became stronger. The
studies that suggest increased risk of dia-
betes with weight loss were based on preva-
lence (17,18) or incidence of diabetes
during the weight change period, and the
increased risk of diabetes in the weight-loss
group is likely to be biased by the inclusion
of diabetic subjects who have lost weight as
a consequence of the disease. Our prospec-
tive design avoids the bias of weight change
after the diagnosis of diabetes.
Weight fluctuation
Long-term maintenance of a reduced body
weight is difficult, and many subjects regain
the lost body weight (40). The possibility
that weight fluctuation may have a diabeto-
genic effect has caused concern about
weight control interventions. The evidence
linking weight fluctuation to increased risk
of diabetes has been assessed in a recent
study that concludes that earlier findings
based mainly on retrospective studies and
one prospective study have limitations and
biases (11). In this study, weight fluctuation
was not associated with increased incidence
of diabetes, and it was concluded that con-
cern about the diabetogenic effect of weight
fluctuation should not deter weight control
efforts (11). Our study provides further sup-
port for the concept that weight mainte-
nance in subjects who are not overweight
and weight loss in overweight and obese
subjects is likely to be effective in the pri-
mary prevention of diabetes.
Public health implications
Obesity is an epidemic in the industrialized
world and is a major risk factor for the
development of diabetes. A recent review of
data from NHANES from 1960 to 1991
showed a progressive increase in mean
BMI, with 1 in 3 adults aged 20–74 years
currently being classified as overweight
(i.e., at least 20% heavier than ideal for
height) (41). The prevalence of diabetes

(diagnosed and undiagnosed) in U.S.
adults aged 40–74 years has increased from
11.4% in 1976–1980 to 14.3% in
1988–1994 (according to World Health
Organization criteria), and this reflects the
gradual increase in diabetes worldwide
(42). It is estimated that, by the year 2025,
the world prevalence of diabetes will be
5.4%, involving some 300 million adults,
and that the major part of this increase will
be in the developing countries and in per-
sons aged 45–64 years rather than persons
aged $65 years as seen in developed coun-
tries (43). Overweight and obesity are the
major environmental threads running
through this epidemic of diabetes. In Great
Britain, nearly half the adult (16–64 years)
population is currently overweight ($25
kg/m2), and the prevalence of obesity ($30
kg/m2) has increased steadily during recent
decades such that 16–17% of British men
and women are now obese (44). Obesity
carries with it a considerably increased risk
of diabetes. It is clear that controlling body
weight requires a population-oriented
approach, and yet the current Health of the
Nation target in the U.K. is the reduction of
the prevalence of obesity (45) rather than a
reduction in the overall distribution of BMI
toward lower and more desirable levels
(37). In the present study, substantial
weight gain (.10%) is associated with a
significant increase in the risk of diabetes in
middle-aged subjects, and weight reduc-
tion in both low- and high-risk subjects is
associated with a lower risk of developing
diabetes. The study emphasizes that the
duration of overweight or obesity is critical
in estimating the risk of developing dia-
betes. The prevention of overweight and
obesity and weight reduction in overweight
or obese people are both likely to be effec-
tive strategies in the primary prevention of
diabetes. Given the poor results of mainte-
nance of weight loss, the prevention of
overweight and obesity at all ages should
be an urgent public health concern.
Acknowledgments — The British Regional
Heart Study is a British Heart Foundation
Research Group and receives support from the
Department of Health. S.G.W. is a British Heart
Foundation Research Fellow.
References
1. Spelsberg A, Manson JE: Towards preven-
tion of non-insulin dependent diabetes mel-
litus. In Causes of Diabetes. Leslie RDG, Ed.
Chichester, U.K., Wiley, 1993, p. 319–345
DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999
2. Rewers M, Hamman RF: Risk factors for
non-insulin-dependent diabetes. In Diabetes
in America. 2nd ed. Harris MI, Cowie CC,
Stern MP, Boyko EJ, Reiber GE, Bennett PH,
Eds. Washington, DC, U.S. Govt. Printing
Office, 1995 (NIH publ. no. 95-1468), p.
179–220
3. Pi-Sunyer X: Weight and non-insulin-
dependent diabetes mellitus. Am J Clin Nutr
63 (Suppl. 3):426S–429S, 1996
4. Tuomilehto J, Knowler WC, Zimmet P: Pri-
mary prevention of non-insulin-dependent
diabetes mellitus. Diabetes Metab Rev
8:339–353, 1992
5. Olefsky JM, Koltermann OG, Scarlett JA:
Insulin action and resistance in obesity and
non-insulin dependent type II diabetes
mellitus. Am J Physiol 243:E15–E30, 1982
6. Reaven GM: Role of insulin resistance in
human disease. Diabetes 37:1595–1607,
1988
7. Henry RR, Wallace P, Olefsky JM: Effects of
weight loss on mechanisms of hypergly-
cemia in obese non-insulin dependent dia-
betes mellitus. Diabetes 35:990–998, 1986
8. Wing RR, Koeske R, Epstein LH, Nowalk
MP, Gooding W, Becker D: Long term
effects of modest weight loss in type II dia-
betic patients. Arch Intern Med 147:
1749–1753, 1987
9. Holbrook TL, Barrett-Connor E, Wingard
DL: The association of lifetime weight and
weight control patterns with diabetes
among men and women in an adult com-
munity. Int J Obes 13:723–729, 1989
10. Chan JM, Rimm EB, Colditz GA, Stampfer
MJ, Willett WC: Obesity, fat distribution
and weight gain as risk factors for clinical
diabetes in men. DiabetesCare 7:961–969,
1994
11. Hanson RL, Narayan KMV, McCance DR,
Pettitt DJ, Jacobsson LTH, Bennett PH,
Knowler WC: Rate of weight gain, weight
fluctuation and incidence of NIDDM. Dia -
betes 43:261–266, 1995
12. Colditz GA, Willett WC, Rotnitzky A, Man-
son JE: Weight gain as a risk factor for clin-
ical diabetes mellitus in women. Ann Intern
Med 122:481–486, 1995
13. Ford ES, Williamson DF, Liu S: Weight
change and diabetes incidence: findings
from a national cohort of US adults. Am J
Epidemiol 146:214–222, 1997
14. Eriksson KF, Lindgarde F: Prevention of
type 2 (non-insulin-dependent) diabetes
mellitus by diet and physical exercise: the
6-year Malmo Feasibility Study. Diabetologia
34:891–898, 1991
15. Long SD, O’Brien K, MacDonald KG Jr,
Leggett-frazier N, Swanson MS, Pories WJ,
Caro JF: Weight loss in severely obese sub-
jects prevents the progression of impaired
glucose tolerance to type II diabetes. Dia -
betes Care 17:372–375, 1994
16. Wing RR, Venditti E, Jakioio JM, Polley BA,
Lang W: Lifestyle intervention in over-
Wannamethee and Shaper
weight individuals with a family history of
diabetes. Diabetes Care 21:350–359, 1998
17. Noppa H: Body weight change in relation
to incidence of ischemic heart disease and
change in risk factors for ischemic heart
disease. Am J Epidemiol 111:693–704, 1980
18. Higgins M, D’Agostino R, Kannel W, Cobb
J: Benefits and adverse effects of weight
loss: observations from the Framingham
Study. Ann Intern Med 119:758–763, 1993
19. Morris RD, Rimm AA: Long-term weight
fluctuation and non-insulin dependent dia-
betes mellitus in white women. Ann Epi -
demiol 2:657–664, 1992
20. Shaper AG, Pocock SJ, Walker M, Cohen
NM, Wale CJ, Thomson AG: British
Regional Heart Study: cardiovascular risk
factors in middle-aged men in 24 towns. Br
Med J 282:179–186, 1981
21. Pocock SJ, Ashby D, Shaper AG, Walker M,
Broughton PMG: Diurnal variations in
serum biochemical and haematological
measurements. J Clin Pathol 42:172–179,
1989
22. Shaper AG, Wannamethee G, Weatherall R:
Physical activity and ischaemic heart dis-
ease in middle-aged British men. Br Heart J
66:384–394, 1991
23. Obesity: A Report of the Royal College of Physi -
cians. London, Royal College of Physicians
of London, 1983
24. Walker MK, Whincup PH, Shaper AG,
Lennon LT, Thomson AG: Validation of
patient recall of doctor-diagnosed heart
attack and stroke: a postal questionnaire
and record review comparison. Am J Epi -
demiol 148:355–361, 1998
25. Lampe FC, Walker M, Lennon L, Whincup
PH, Ebrahim S: Validity of self-reported
history of doctor-diagnosed angina. J Clin
Epidemiol 52:73–81, 1999
26. Wannamethee G, Shaper AG: Weight
change in middle-aged British men: impli-
cations for health. Eur J Clin Nutr
44:133–142, 1990
27. Walker M, Shaper AG: Follow-up of sub-
jects in prospective studies in general prac-
tice. J Royal Coll Gen Pract 34:365–370,
1984
28. Cox DR: Regression models and life
tables. J Royal Stat Soc 34 (Sect. B):87–
220, 1972
29. Walker M, Wannamethee G, Shaper AG,
Whincup PH: Weight change and risk of
coronary heart disease in the British
Regional Heart Study. Int J Epidemiol
24:694–703, 1995
30. Manson JE, Nathan DM, Krolewski AS,
Stampfer MJ, Willett WC, Hennekens CH:
A prospective study of exercise and inci-
dence of diabetes among US male physi-
cians. JAMA 268:63–67, 1992
31. Perry IJ, Wannamethee SG, Walker M,
Thomson AG, Whincup PH, Shaper AG:
Prospective study of risk factors for devel-
opment of non-insulin dependent diabetes
1271
Weight change, duration of overweight, and diabetes
in middle-aged British men. BMJ 310:
560–564, 1995
32. Everhart JE, Pettitt DJ, Bennett PH, Knowler
WC: Duration of obesity increases the inci-
dence of NIDDM. Diabetes 41:235–240,
1992
33. Colditz GA, Willett WC, Stampfer MJ, Man-
son JE, Hennekens CH, Arky RA, Speizer
FE: Weight as a risk factor for clinical dia-
betes in women. Am J Epidemiol 2:501–513,
1990
34. Swinburn BA, Nyomba BL, Saad MF, Zurlo
F, Raz I, Knowler WC, Lillioja S, Bogardus C,
Ravussin E: Insulin resistance associated
with lower rates of weight gain in Pima Indi-
ans. J Clin Invest 88:168–173, 1991
35. Berger M, Bannhoff E, Gries GA: Effect of
weight reduction on glucose tolerance in
obesity: a follow-up study of five years. In
Recent Advances of Obesity Research. Harvard
A, Ed. London, Newman, 1975, p. 128–133
36. American Health Foundation Roundtable
1272
on Healthy Weight. Am J Clin Nutr 63
(Suppl. 3):S409–S477, 1996
37. Shaper AG, Wannamethee SG, Walker M:
Body weight: implications for the preven-
tion of coronary heart disease, stroke and
diabetes mellitus in a cohort study of mid-
dle aged men. BMJ 314:1311–1317, 1997
38. Henry RR, Gumbiner B: Benefits and limi-
tation of very low calorie diet therapy in
obese NIDDM. Diabetes Care 14:802–823,
1991
39. Wing RR, Koeske R, Epstein LH, Nowalk
MP, Gooding W, Becker D: Long term effects
of modest weight loss in type II diabetic
patients. Arch Intern Med 147:1749– 1753,
1987
40. Kramer FM, Jeffery RW, Forster JL, Snell
MK: Long-term follow-up of behavioural
treatment for obesity: patterns of weight
regain among men and women. Int J Obes
13:123–136, 1989
41. Kuczmarski RJ, Flegel KM, Campbell SM,
DIABETES CARE, VOLUME 22, NUMBER 8, AUGUST 1999
Johnson CL: Increasing prevalence of over-
weight amongst US adults: the National
Health and Nutrition Examination Surveys
1960–1991. JAMA 270:205–211, 1994
42. King H, Aubert RE, Herman WH: Global
burden of diabetes, 1995–2025: preva-
lence, numerical estimates, and projections.
Diabetes Care 21:1414–1431, 1998
43. Harris ML, Flegal KM, Cowie CC, Eber-
hardt MS, Goldstein DE, Little RR, Wied-
meyer HM, Byrd-Holt DD: Prevalence of
diabetes, impaired fasting glucose, and
impaired glucose tolerance in US adults:
the Third National Health and Nutrition
Examination Survey, 1988–1994. Diabetes
Care 21:518– 524, 1998
44. Bennett N, Dodd T, Flatley J, Freeths S,
Bolling K: Health Survey of England 1993.
London, HMSO, 1993
45. Department of Health: The Health of the
Nation: A Strategy for Health in England.
London, HMSO, 1992